Introduction

​

Welcome to Agents Squads

We build transparent AI systems that teams can learn, understand, and own — not black boxes, but reliable systems earning trust through consistent results.

​

Why Agents Squads?

Trust is the bottleneck, not capability. Most AI systems fail not because they can’t perform — but because teams can’t understand, modify, or trust them. We take a different approach:

• Transparent — See exactly what agents do and why
• Learnable — Your team can understand and modify the system
• Ownable — No vendor lock-in, you control your agents
• Reliable — Consistent results you can depend on

​

How It Works

Agents Squads organizes AI capabilities into domain-aligned teams — and closes the loop: every cycle’s output is evaluated and fed back into the next. squads init ships four starter squads:

​

The execution loop

Every run moves through the same four zones. Context: the run assembles only the minimal slice its task needs — agent definition, priorities, relevant squad memory, and typed views over your data. Execution cycle: the agent reasons, calls scoped schema-checked tools, and renders its output for review. Human in the loop: you approve, reject, edit, or escalate — nothing commits without the gate. Outcomes: external work product lands (data writes, artifacts) and internal memory accumulates (state, learnings) — so the next cycle starts richer. See Architecture for how the underlying context cascade works.